Unofficial Opteron 939 Thread

[Squid1]

Thanks Martin but I should have been asking for CPU VID multiplier.

ie, 104%, 110%, 133%, 126%, etc.

or voltage readings at 1.5 vs 1.2x126%

trying to understand if the voltage error is different using the VID multiplier.

(long day at work, hope I don't sound more brain dead than usual)

 

"ix" is right on target with the measuring and using Autchthon's calibration post.

links to EMC2's voltage test points are also in Post #2 of the MBM5 guide if ya care

 

these ARE really small voltages and the meter itself make have some error depending on quality.

 

For the Opteron I only used the primary voltage control, no special control.

 

With my Venice I used special control but couldn't find a difference between straight volts and with special control (provided <= 1.55 volts). Didn't try very hard, though. I supposed I can try it on the Opteron.

[omi7000]

hi all... my 148 oem from Monarch arrived..

 

it's a..........

 

CABYE 0535XPMW ..I'm building a new rig, wont have it complete until the weekend. Not sure how much i can push it on air with my SI-120. Loking into wc'ng real soon.

 

EDIT: correction on stepping spelling.. Sorry!

[ixtapalapaquetl]

Here are some initial results regarding Vcore:

 

BIOS VID setting: 1.45*1.10 = 1.595V

BIOS reported Vcore: 1.570V

BIOS measured Vcore: 1.616V

 

CPUZ load/idle VCORE: 1.552V

Windows idle measured Vcore: 1.609V

Windows load measured Vcore: 1.659V

 

 

It is this last reading that is especially troublesome - we all know these boards undervolt w.r.t. monitoring, but this massive .05V jump under load does not manifest itself under any software monitoring program I've tried (CPUZ, MBM, SpeedFan, OCCT...)! I'm so glad I happened to decide to finally check out the nF4 voltages (only had it like 2 weeks) while priming!!!

 

More info to come.

[akin]

OK got my Opty 175 running stable on my DFI NF3 Ultra D took some effort but Primed for 13 hours....

 

http://users.on.net/~akin/27013.PNG

 

Link to screenie when finished the thing got to 43c temp wise on Air which I am not that unhappy with...I will leave for a week and see how high I can get but so far the Voltage is getting scary 1.575 Vcore to get here

 

Had to crank up the 12volt rail on the psu as well now showing 12.30 in BIOS which is a tad high I think....any advice would be appreiciated to achieve a higher clock

 

Cheers all

[My Ugly PC]

hi all... my 148 oem from Monarch arrived..

 

it's a..........

 

CABEY 0535XPMW ..I'm building a new rig, wont have it complete until the weekend. Not sure how much i can push it on air with my SI-120. Loking into wc'ng real soon.

 

Did you accidentally switch the E and Y?

 

Windows load measured Vcore: 1.659V

 

And, ix, you got that multimeter for a while? Maybe you could test different setups, or maybe even the NF3. And if I were you I wouldn't prime for crazy amount of hours anymore, after seing your load voltage. I'm sure like 4 hours stable is plenty good, I've never primed overnite or when I'm at work.

[ixtapalapaquetl]

Multimeter borrowed from Electrical Engineering friend, so it's known to be solid. Craftsman Professional #82325.

 

Here are some more results similar to initial reports. Additionally, as per Soundx98's request, you can see that the VID multipliers are not terribly accurate. Each row below should be very similar since they are all very near 1.5V. This does not pan out, however.

 

 

I think the moral of the story is simply to exercise caution. My chip told me it was running high before I measured it; I would have not gone higher even if I didn't have this data. But please know that your load voltages may be significantly higher than what CPUZ tells you, perhaps even more than previously reported. I would love to know if anyone can confirm/reject these readings on their own rigs. A thought I recently had is that Opterons, known to clock well at low voltage, may not be doing so in reality! We could just be getting a phantom Vcore boost while testing stability! I believe the IRS calls this unreported income - and the penalty is high if you get caught.

 

Again, let me know if anyone can reproduce any of this data.

 

Thanks.

[soundx98]

Thanks for the effort "ix"

 

I think we were all aware that cpu-z reports voltages lower than actual. When using OCCT I was quite surprized that the Vcore increased under load as I would have assumed it would sag lower.

 

 

your results seem to be verifying my assumption that a) the vid multipliers are inaccurate B) Vcore actually increases under load.

 

Very wierd that all voltages would not sag slightly under load as they normally due on an electronic device.

[EdLSmith]

I obtained an el cheapo MM from Ebay and will set up to check voltages today.

Looks like the Voltages reported are fairly close until the chip goes under load and then there is a significant increase of about .05 Volts. That is .05/1.51; about 3.5%. Yes, I will definitely check Voltages.

 

Now the status on my Opteron 148 CABGE 0528. I reached 270 on Vcore of 1.40X104%

and then when I tried to go to 273, it showed some instability. I then jumped vcore to

1.40 X 113% or 1.58 computed, which I can estimate as 1.62-1.65 following ixtapalapaquetl's measurements as a guide to estimate by. This Vcore would appear to be my limit for sanity. I'm running SuperPi 32m right now and then I will run OCCT, followed by 8 hour Prime. Hope it's stable, but it looks good so far.

[ncsa]

Here are my measurement checks of the Vcore.

 

 

I do not see the large variation which 'ix' is experiencing and while CPUz is reporting low, the difference is small on my particular MB. It would seem that the VID Settings are very close to the measured Vcore.

 

If in doubt then check it out, as it could be that the findings for an individual's setup might be surprising ...

 

A very worthwhile exercise to do and has helped confirm that this Opteron is performing very well at a nice low Vcore. :nod:

[EdLSmith]

Here is my FX and Opteron:

 

Neat, what?

[soundx98]

I was waiting for someone to post the "right on" readings, thanks my friend.

As I said very small voltages to measure. and of course variables between DMMs and boards and calibration to consider.

ncsa, any chance we can ask you to measure some higher VID multipliers as well (126%, or 133%)

[tatts]

do you think its possible that the vcore is rising under load due to power supply?

i run mine on stock volts 2.8ghz to test and is rock steady, will test again later with 1.5V see if its still a straight line